Reliability: unsaved rows can no longer be lost on restart (rows travel with their detail request), one writer saves then charges each row once, soft deadline ends long runs as PARTIAL, honest outcome (COMPLETE, PARTIAL, VALID_EMPTY, UPSTREAM_FAILED) in OUTPUT and RUN_SUMMARY with a plain-language agentSummary.
Cost: datacenter proxy first, residential only when blocked.
New: optional organizer profiles (followers, hosting years, events hosted, website and socials), list-only rows, detailsFetched flag, free-plan cap of 200 events per run.
Fix: organizer page URL no longer lands in organizerWebsite.
Reliability: when a browse page returns 200 but is missing SERVER_DATA (transient proxy issue or partial CDN response), retire the session and force a retry instead of silently dropping the page. Bump maxRequestRetries 5 -> 8.
Fix free-text search URL. Eventbrite redirects /?q=foo to // but the redirect happens via 301 the embedded SERVER_DATA is only on the canonical slug URL. buildSearchUrl now produces /d/// when only searchQuery is provided.
Fix dataset schema validation for nullable numeric fields (venueLatitude/Longitude, priceMin/Max, imageWidth/Height). Online events and RSVP-only events now push correctly.
Initial release. Three input modes, embedded SERVER_DATA extraction, Schema.org JSON-LD enrichment, residential proxy default.